[solotronics]

Do you think that with integrating AI/ML techniques fully into manufacturing automation that in our lifetimes people will become fully obsolete for this kind of work? As a cloud/software/network guy I am curious to see someones opinion who is knowledgeable in this area.

[wigiv]

Within our lifetimes (say, the next 40-60 years), no, personally I don't think we'll see completely autonomous end-to-end manufacturing widely implemented (as much as I'd like to, considering it's a problem space I focus on!).

Some pockets of industry are much further ahead than others, but it will take A LOT of work to reach parity across the board. If not for technical reasons (which I'm more optimistic about), then for political and social reasons, as these systems and understandings adapt. That's a whole 'nother discussion...

AI/ML will play a huge role. Not only in machine resilience once commissioned and operating, but upstream and downstream as well. Better (AI/ML-assisted) tools for designing products and the factories/equipment that make them will preempt some of the challenges caused by the currently disjointed process.

I disagree with the comment that AI/ML techniques are only useful once you've physically built a plant - there are of course emergent behaviors that only crop up when dynamics of the whole unique factory are at play, but any given problem that arises is almost always traceable to one or a small number of subcomponent failures, for which better, more granular datasets are becoming available to train AI upon.

And, as I mentioned in my comment about throwing virtual wrenches in virtual works - simulations can begin to generate training data sets as well!

[jeffreygoesto]

They won't. A new line needs to be "trained" for a ML system to be useful. In our lifetime I don't expect generalization across manufacturing lines to be feasible or economically advantageous.